Tree-loss function for training neural networks on weakly-labelled datasetsSergey DemyanovRajib Chakravortyet al.2017ISBI 2017
